Active Object Localization with Deep Reinforcement Learning

By Juan C. Caicedo, Svetlana Lazebnik. IEEE International Conference on Computer Vision, 2015

468 citations recorded by the source corpus.

We present an active detection model for localizing objects in scenes. The model is class-specific and allows an agent to focus attention on candidate regions for identifying the correct location of a target object. This agent learns to deform a bounding box using simple transformation actions, with the goal of determining the most specific location of target objects following top-down reasoning. The proposed localization agent is trained using deep reinforcement learning, and evaluated on the Pascal VOC 2007 dataset. We show that agents guided by the proposed model are able to localize a single instance of an object after analyzing only between 11 and 25 regions in an image, and obtain the best detection results among systems that do not use object proposals for object localization.

Subfields: Aerospace Engineering, Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/2735e6e760a5fce119d82dc1e449f329d072e172

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

An Artificial Agent for Robust Image Registration (2016, methodology): “In (Caicedo and Lazebnik 2015) an active detection model for localizing objects in 2-”
An Artificial Agent for Anatomical Landmark Detection in Medical Images (2016, background): “In parallel to our work, similar ideas have been exploited also in the context of 2D object detection [2].”
Attention to Scale: Scale-Aware Semantic Image Segmentation (2015, methodology): “Attention models for deep networks: In computer vision, attention models have been used widely used for image classification [8, 25, 53] and object detection [4, 7, 57].”
Deep Reinforcement Learning for Active Breast Lesion Detection from DCE-MRI (2017, methodology): “0 used in [8] helped triggering when finding a lesion).”
Semantic Image Retrieval via Active Grounding of Visual Situations (2017, background): “More recently, several groups have framed active object detection as a Markov decision process and use reinforcement learning to learn a search policy (e.g., [27]).”
Part-Activated Deep Reinforcement Learning for Action Prediction (2018, background): “[1] defined a transformations set for the bounding box as the action of agent and rewarded the agent when the bounding box moves close to the ground-truth with iterations.”
Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning (2017, methodology): “Also, several methods were proposed to solve the computer vision problems, such as object localization [3] or action recognition [16], by employing the deep RL algorithms.”
On the Detection of Digital Face Manipulation (2019, methodology): “In computer vision, attention models have been widely used for image classification [12, 51, 56], image inpainting [33, 60] and object detection [11, 59].”
Weakly Supervised Object Localization and Detection: A Survey (2021, unknown): “Several existing methods [11], [63], [73], [94], [187] have also demonstrated that designing deep reinforcement learning frameworks to model such a sequential searching process can indeed help to address the object localization, detection, and tracking problems in the computer vision community.”
Spatial Memory for Context Reasoning in Object Detection (2017, background): “An interesting recent direction focuses on using deep reinforcement learning (DRL) to optimize the sequence selection problem in detection [6, 9, 50, 61].”
Learning to Look Around: Intelligently Exploring Unseen Environments for Unknown Tasks (2017, background): “While there is interesting recent headway in active object recognition [3, 11, 29, 43] and intelligent search mechanisms for detection [10, 32, 45, 73], such systems are supervised and task-specific—limited to accelerating a predefined recognition task.”
An analysis of rotation matrix and colour constancy data augmentation in classifying images of animals (2018, unknown): “Another interesting study is the use of deep reinforcement learning for active localization of cows (Caicedo & Lazebnik, 2015).”
Reinforcement Learning for Visual Object Detection (2016, methodology): “The stochastic policy is given by a Gaussian distribution on top of highly non-linear predictions (in contrast notice that methodologies like [14] [4] are deterministic).”
Reinforced Attention for Few-Shot Learning and Beyond (2021, background): “The applications are also studied on more classic tasks, such as classification [61], object detection [2, 35], segmentation [1, 66] and pose estimation [24, 44].”
MathDQN: Solving Arithmetic Word Problems via Deep Reinforcement Learning (2018, background): “…is that deep Q-network has witnessed success in solving various problems with big search space such as playing text-based games (Narasimhan et al. 2015), information extraction (Narasimhan et al. 2016), text generation (Guo 2015) and object detection in images (Caicedo and Lazebnik 2015).”
Attend Refine Repeat: Active Box Proposal Generation via In-Out Localization (2016, unknown): “Note that active localization schemes have also been previously applied in the object detection literature [4, 13, 14, 17, 30, 39, 42].”
Evaluating Reinforcement Learning Agents for Anatomical Landmark Detection (2019, unknown): “One solution is to define a new trigger action that terminates the sequence of the current search when the target state is reached [28, 39].”
Tree-Structured Reinforcement Learning for Sequential Object Localization (2016, background): “It does not rely on running multiple episodes starting from the whole image like [19] to find multiple objects.”
Sketch-Guided Object Localization in Natural Images (2020, background): “ed. In [3,31], the authors proposed attention networks to generate attention scores on each of the object proposals, and an attention-based dropout layer for weakly supervised object localization. In [2] a dynamic attention-action strategy using deep reinforcement learning was proposed to localize o”